Perceptions Research for Stability & Countering Violent Extremism (PRSCVE)
Perceptions Analysis, 2014-2015 in Nigeria (abridged)
October, 2016
Data | Context | Impact
Program Introduction
Perception Research for Stability & Countering Violent Extremism (PRSCVE), pronounced “Perceive”, is Phase Zerφ’s indigenous field-based persistent impact monitoring & evaluation instrument that amplifies perception and sentiment of local communities for a deep contextual understanding of drivers & indicators of instability and violent extremism.
PRSCVE simultaneously develops community baselines while persistently monitoring rapidly evolving security, governance, economic, development, political, social, cultural, religious, civil, and human rights issues in their local context
PRSCVE provides rich locally-nuanced content to design influential population engagement strategies which, once applied, are then monitored and measured for impact providing a deep learning system of behaviors and responses in the operational environment
PRSCVE has previously supported critical Department of Defense (DoD), Department of State (DoS), and United States Agency for International Development (USAID) programs to include counter-terrorism, capacity building, force protection, key leader engagements, counter-narratives, strategic communications/messaging, and information operations
Data Description and Context
Over 2,200 individual observations were used in this product covering the period of November 2014 - April 2015 which corresponds to the rise of Boko Haram, establishment & collapse of their Islamic Caliphate, and through the country’s Presidential election. This interactive HyperText Markup Language (HTML) product highlights examples of ways to visualize the near real-time data from our field-based PRSCVE network and is not a focused analytic product itself. PRSCVE captures detailed information on population demographics, language, local knowledge, sociocultural context, local researcher comments, and detailed geographically referenced perceptions and sentiments of local populace. This information is completely unclassified and should be shared with the U.S., Multi-National Partner and host nation community.
Long-term Trends: Buhari and Jonathan Sentiment
Each observation in the field includes a sentiment score that subjectively measures resident’s sentiments toward a theme or entity. Here we plot sentiment toward President Goodluck Jonathan and General Buhari before, during, and after the election to see how resident’s perceptions of them changed. Sentiment toward Boko Haram is also plotted. The vertical dashed lines represent significant events. The slider beneath the graph can zoom in on your desired date ranges. The number in the box in the bottom left corner controls line smoothing.
After the Baga attacks in January, sentiment toward General Buhari dropped from highly positive to neutral while President Jonathan’s slightly increased. After February these trends reversed and people began talking more positively about Buhari and negatively about Jonathan as the election neared.
Because our observers recorded sentiment scores for the people and entities mentioned in the narrative reports, these kinds of trends are possible for many entities. For best results, compare entities that have many mentions across time.
A report from March 24th, 2015 provides insight into specific sentiments about President Goodluck Jonathan:
The 1st man in his 40’s said “Allah ya tsine Jonathan,” meaning, “Allah should deal treacherously with the President of Nigeria, Jonathan, for he is the cause of dwindling in their economic fortune. The second man interjected,”Allah ya kashe shi dai," meaning, “Allah should kill Jonathan.” He reeled out how their daily take-home pay can hardly take them home because of the activities of Boko Haram that has seriously affected the number of trucks they load. The 3rd man, however, objected, saying, “Killing Jonathan is not a good prayer but rather, they should pray that Allah should change Jonathan’s heart if he is destined that way but if he is not destined to change, then Allah should do away with him.”
The same report also provides details about the location:
Bolori Store is a hub of commercial activities known for transporting goods to some African countries especially Tchad, Niger Republic, Cameroon and as far as Central African Republic by road. However, in the wake of the insurgency and the closure of borders, this once thriving wholesale business enterprise has lost its known glory. To add salt to injury, merchants have only one route to take goods out of Maiduguri, the Maiduguri-Bauchi Road, as a result of insurgents’ activities. The store boasts of over 50 workers ranging from the merchants and real traders, store keepers/attendants, laborers/loaders, drivers of over 20 trucks loaded each day when business was thriving, food sellers and hawkers of consumables. On this day, NIC0010 heard 3 low loaders of the store lamenting their plight.
Changing Sentiments
How has sentiment toward top entities changed?
People felt more positive about Gen. Buhari than any other entity. People were pretty neutral about Goodluck Jonathan, but sentiment towards him had been increasing. Unfortunately, people were neutral about Boko Haram, though sentiment towards them was increasing.
The horizontal orange line is the mean of all sentiment scores. In this plot sentiment scores were normalized then averaged.
How do Other Sentiments Vary by State?
To illustrate another way of looking at sentiments across states, we can use a tile plot. Here, we are only looking into states where we had at least 100 reports, and only at entities that were mentioned at least 17 times.
In Yobe, observed residents felt positively about just one thing: General Buhari. In Kano, sentiments were all negative. Looking across rows, President Goodluck Jonathan edged into positive territory only in Benue. The Nigerian Police and federal government were relatively unpopular in all of these states. General Buhari was positive across the board, but the elections themselves were viewed quite negatively in FCT and Kano.
Themes and Sub-themes in the narratives
Streamgraphs represent the bandwidth of reporting across time. The vertical width of the entire stream represents the amount of reporting at any given time. For example, we had more reports in December than any other month. The individual colors within the stream represent the amount of reporting on specific themes. Hover your mouse over each stream to see the theme and its count. Click on each tab to see different streamgraphs.
Streamgraph for theme 1 (by month)
Streams representing first-level themes. Across time, governance was the most mentioned theme, followed by security and social themes. Development and economic themes received less reporting.
Streamgraph for all sub-themes (by month)
Streams representing all sub-themes. Across time, elections and political participation were the most mentioned sub-themes. Religion and corruption were also consistently mentioned sub-themes.
Streamgraph for sub-themes of governance (by month)
We can dig into specific themes for more detail. These streams represent all sub-themes under the governance main theme. Across time, elections and political participation were the most mentioned sub-themes. Corruption appears to be the next most mentioned sub-theme, followed by local leaders and power brokers. The spikes are also informative. In January and towards the end of March reporting on elections and political participation spiked as the national elections approached.
Demographics
Who is in our data?
Demographic proportions in our observations:
- 26 percent female, 74 percent male
- 48 percent lower class, 48 percent middle class, 4 percent upper class
- 91 percent wore secular dress, 9 percent religious dress
Age
The age distribution of our observations is more easily visualized in a violin plot.
The plot shows the age distribution of local residents whose sentiments were captured in our reports. The width of the colored areas widen or narrow depending on how many residents were that age. The dot in each colored area is the average age.
The average female age was 37, and the average male age was 39. Approximately 82 percent of people observed were between 20-49 years old. In the field, exact ages were not recorded, but were instead recorded in categories such as 20-29, 30-39, etc. For plotting purposes, we re-coded each observation to be the middle of its category. Everyone in the 20-29 bucket was re-coded to be 25. Everyone in 30-39 bucket was re-coded to be 35. And so on.
Language
One final demographic to show before digging into the rich narratives is the language distribution among our data. Here we show a table of languages spoken in each state where we had at least 100 reports. Those states are Kano(108), Plateau(112), Borno(166), Yobe(464), Benue(612), Adamawa(740).
Sentiments by demographic
Now that we know who is in our data, we look at how sentiments are different among the demographic groups. We will look into just one example here, though many more are possible.
In Kano, the difference between the average male and female sentiments is greatest. Both are negative, but female sentiment is much more negative. This observation leads to further investigation. We can look just within Kano and see how sentiments differ by reporting theme.
This table shows that within the state of Kano, male and female sentiments differ significantly on the theme of governance. Across all the other themes, female sentiment is most negative, too. Because there are fewer reports from females, their average sentiments will tend to be more extreme than the males whose average sentiment is moderated by the large number of male reports. However, because female sentiments are consistently negative across multiple themes, we can investigate further. At this point we look at the narratives in the original reports to see how women in Kano are expressing their negative sentiments.
From the original reports here we find the central topic of discussion is the town of Sabon Gari. Our reporter describes the area:
Sabon Gari is very unique in that it is the only place in Kano state that can boast of accommodating all calibre of Nigerians. So it is a multicultural settlement including foreigners. Apart from the vices attributed to Sabon-Gari like brothels and drugs, Sabon-Gari also boasts of decent houses of Christian worship numbering over a thousand. Virtually every street in Sabon-Gari has a church or a prayer house. In addition to churches, it is the only place that has accommodated the two big religions of Nigeria (Christianity and Islam), there are mosques and churches facing each other or built side by side and there is hardly any problem.
Our reporter had a conversation with “Madam Catherine, a popular Igbo woman selling Pepper Soup (a special Nigerian delicacy of intestines cooked with lots of pepper and served as soup, the same soup can be prepared using cowtail, head and legs).” She reported that:
…there is no more business. Most people have either relocated from Kano or have taken their families away because of the upcoming elections, nobody wants to stay for fear of potential violence come the 14 February.
Our reporter then listened to a conversation between two women in the Yankura Market in Sabon Gari:
They both complained that Rabiu Musa Kwankwaso was massively voted for in Sabon-Gari by the people who came out in their numbers to vote for him. But according to one of the speakers, “Kwankwaso has betrayed our trust, first he decamped from People’s Democratic Party (PDP) and then deceived all of us by starting to repair the new road that is very important for the people and then later he abandoned the project.” “This type of behavior,” said the other women, “is very very bad…imagine this man with all his experience as former Defence Minister O! Well Na God go help us sha…”
And from a third woman, our reporter writes:
On the same street is Miss Esther a young woman who had been roasting fish for sale in front of the Hotel La Mirage who lamented that her business is crippling and that there are no customers, she said, sometimes she could sell more than 50 pieces of fish but now she barely sells 10 or 15.
We hear from these women deep concerns about election violence and failed government promises regarding development of infrastructure. One woman expressed a sentiment of betrayal because she felt the people of Sabon Gari had done so much work to elect Kwanwaso who then failed to complete the road project.
These sentiments also show that these women’s concerns are very local and focused mostly on how government issues are negatively affecting their personal safety and ability to earn income.
Mapped reports by theme
Perceptions reports are clustered under the colored circles which will de-cluster as you click on them to zoom in. Black dots are not clustered; they represent significant activity and are sized according to number of fatalities per event. All symbols will show a pop-up of report information when you click them.
| Fatalities | Per Group |
|---|---|
| under 6 | 867 |
| 6 - 27 | 174 |
| 27 - 59 | 40 |
| 59 - 126 | 24 |
| 126 - 230 | 5 |
| over 230 | 6 |
Points of contact:
Mike Williams, Founder and CEO mikewilliams@phasezero.io
Jeremy Allen, Junior Data Scientist jeremya@phasezero.io
Phase Zero
6295 Greenwood Plaza Blvd.
Greenwood Village, CO 80111-4908
Phone +1 (855) 474-2730 x700
https://phasezero.io
info@phasezero.io
Social Class
Using the same kind of plot, we look at social class as a function of age. We limit this plot to only states where we had more than 100 reports
Again, the dark dot in each violin is the average age per social class in each state. Take Adamawa for example. Among the people our reporters observed, the average age of lower-class residents was 32.6, for middle class 40.7, and for upper class 52.7. The trend, which is visible in each state, is that higher social class also indicates higher age, on average. This is very present in our observations in Yobe, where all upper class residents observed by our reporters were over 50 years of age.
As mentioned above, the width of each violin plot indicates the amount of observations at that age. For example, in Kano, the plot really widens for middle class people at 45 years of age, indicating our reporters observed more middle-aged people in their forties than other ages. In Plateau, most lower-class residents observed by our reporters were in their 20s.
It’s important to note that our raw data is not representative of the actual demographic ratios in Nigeria. For example, the gender ratio in Nigeria is approximately 51 percent female and 49 percent male, so women are underrepresented in our observations because the ratio in our data is 26 percent female, 74 percent male. This matters more for quantitative analysis where aggregate measures would be influenced by over and under representation.
The value of our data is that it includes the full text of observations on the ground, so rich qualitative information can be drawn from the narratives to shed light on how residents feel, what they think about various topics, how they relate topics together, and how they think through complicated ideas and concepts around security, governance, development, etc., all of which is opaque to analysis that is solely quantitative.